ABSTRACT
The estimation of health risks is most commonly performed using regression models with the aim of developing a realistic model of the system in question, to identify the variables of interest and to estimate the strength of their effect on a health outcome. Though an epidemiological regression model cannot itself prove causality (that can only really be provided by randomised experiments), it can indicate the change in the response variable that might be associated with changes in exposure which is a very useful tool in understanding and developing insight into possible causal relationships.
